Please enable JavaScript.
Coggle requires JavaScript to display documents.
Research Methods & Statistics - Coggle Diagram
Research Methods & Statistics
V. & VI. INFERENTIAL STATISTICS & STATISTIC TESTS
Inferential Statistics
-> helps us
make informed judgments or estimates about a population
by studying a smaller subset of that population.
2 types of Inferential statistics ->
Parametric vs. Non-parametric
Non-parametric
-> "NO" is used to analyze
N
ominal or
O
rdinal data
Chi-squared
= Mr. Independent = 1 variable only (e.g. persons' gender)
single sample chi-square
= goodness of fit
count all variables to determine the # of variables for a chi-square test
multiple sample chi-square
= used for contingency tables
use when there are 2 variables (e.g. gender (male/female) AND voter preference (democrat/republican)
Chi-Square is a
test of difference
-> it requires the following core conditions
Kai Rasporivich is very independent from other families b/c they are weirdos
i).
randomly selected
from the population
ii). must have
independent observations
-> so we can't measure it more than once
iii). it uses
nominal (categorical) data
Advantages
not susceptible to outliers
easier to calculate
not as powerful as parametric tests
greater chance of type II errors
use if the data meets one of the following criteria
ordinal data
-> has order (e.g. Likert scale or rank)
nominal data
gender (male/female), marital status (single, married), etc.
Non-parametic means -> that
the numbers CAN'T be used in calculations
non-parametic data is a label
Substitute "variable" with "sample"
Parametric
(Para is powerful)
Types of parametric
-> means that they all have
"parameters" or requirements
T-Test = "t" for 2 groups ->
is used to compare 2 means "averages"
i).
t-test for single sample
-> comparing an obtained
sample mean to a population mean
(e.g. mean mock EPPP scores vs. actual EPPP scores)
(e.g. comparing the achievement scores of sixth-grade students in one school district in California to the test scores of all sixth-grade students in California)
iii).
related samples t-test
-> comparing 2 groups when there is relationship btwn them (e.g. twins)
ii).
unrelated t-test
-> difference btwn groups that are unrelated
2 scoops = 2 groups
Use when you have 1 IV and 2+ groups, you are not able to use t-tests it must be a t-test
(3) core requirements of parametric
i).
interval/ratio data,
"ir-data" which has a score value (e.g. income, IQ)
iii).
homoscedascity
= equal varaiability between the groups, which is measured by SD
ii).
normally distributed data
(e.g. bell shape)
If these criteria are not met, you must do a non-parametric tests
"homo" sapien
"ir"ritating sound = interval/ratio
Parametric means that ->
the numbers can be used in calculations
Substitute "variable" with "sample"
Once the data is collected we need to determine which tests to run
Tests of Difference which includes ANOVAS, t-tests & Chi-Square
Factorial ANOVA
-> "factorial" is a generic term that means more than 1 I.V.
"way" actually means IV's -> 2 IV's = 2-way = 3 IV's and 3-way
more than 1 IV -> (2 IV's = 2-way ANOVA; 3 IV's = 3-way ANOVA, etc.)
hallmark is all IV's are correlated
factorial assumes there is more than 1 IV
1-way ANOVA
-> "a" is singular
1-way ANOVA produces an "f-ratio"
F-ratio
**memorize this
-> the larger the f-ratio, the more we can be sure that the changes are due to types of tx -> f-ratios can be as high as 2.0
mean within (MSW)
-> measure due to
error only
mean between (MSB)
-> measure of variability in D.V. that's due to
tx effect plus error
1-way ANOVA = 1 IV and 1 DV
conduct a post-hoc (after the fact) only if statistically significant
MANOVA
-> (Multi-variate analysis of variance) -> "m" multiuple factors
1+ IV
If the Question has more than 1 DV's = choose MANOVA
& get the fuck out of the question
1+ group
For example -> effects of exercise on weight and stress level
ANCOVA
used to ->
statistically remove the effects of an extraneous variable from scores on the D.V.
When using the ANCOVA, the
extraneous or moderator variable
is the covariate
alternative verbage that might be used on the exam for ANCOVA is
hold constant
OR
partial out the effects of
example -> study the effects of smoking on diet and accounting for anxiety (anxiety is extraneous variable )
ancova erases
Trend Analysis
-> use when there is
1 or more IV's
used when the researcher wants to determine
if there is a significant linear or non-linear relationship
A trend analysis shows the ups and downs in a set of data.
For example
nonlinear trend
-> a lower dose has a minimal or even negative effect
a
liner trend
might be -> the higher the dose, the more effective the medication
tires only run in a linear fashion
Split-plot ANOVA
2 IV's -> Independent on 1 variable AND correlated on another variable
Step #1
-> in what form is the data? you do this by looking at the DV (dv=outcome variable)
Step #4 -> is the data (groups) correlated or independent
-> 3 ways to do this
b). if people in groups are related to each other (e.g. twin studies)
c). match people in pairs before you put them into groups
a).
measure people over time
(e.g. beg/middle and end of tx) -> this is the
most common
->
time is always correlated data
Step #3 -> how many IV's?
(the way the groups are being compared)
and what are the levels
(e.g. study comparing medications in treating depression; SSRI, SNRI, MAOI, TCA => 1 variables, with 4 levels)
independent data, is that you can't be in 2 group
Step #2 -> Is the data Interval or ratio?
NO
by default, the data must be either Nominal or Ordinal
conduct a Non-parametric test (e.g. Chi-Square)
YES
data must be parametric,
and you would do the next steps
In order to run a parametric test you must meet 2 additional criteria
i).
homoscedasticity
"homo" = same
scedasticity = variability, or spread (same standard deviation)
ii). normal distribution of the data
Null Hypothesis
-> in layman terms states that....
there is no difference between my control group and my experiment group
Null hypothesis = not happening
I have a theory that what I am going to do is not going to work and I am going to disprove it
Essentially there is NO difference between my control group and my experimental group
We always state our hypothesis in terms of the "null hypothesis" -> we use this language to prevent researcher bias
Next, we create a hypothesis to "reject the null" b/c we are hoping that there will be some differences => rejecting the null is a good thing
Type 1 and type II errors
type II error
(beta) -> I say that my treatment doesn't work but it actually did work
type I error
(alpha) -> I say that my treatment works, but it doesn't
Significance of the research is p-value, 2 options;
alpha is .05 -> which means that there is 95% change that the differences were not coincidental
Beta is .01-> which means there is a 99% chance that the differences is not coincidental
CJ says ->
If the null is rejected,
and this decision later turns out to be a mistake
This is a type I error.
In other words, I found significance in the original experiment => but subsequent researchers don't find significance
CJ says
-> there's no difference between the 2 groups, but other people do the exact same experiment and find out that there are differences -> I said my tx didn't work when it actually did work -> So significance is found in my experiment and I have correctly rejected the null.
https://www.youtube.com/watch?v=SBt7q2m_Ncw
I.
Types of Variables & Data
Variables
-> they are "placeholders"** or buckets for information
Mediator/Intermediate variable
-> helps to explain how or why the IV affects the dependent variable (e.g. if
, then
)
fitness level as a mediator to improved cardiovascular health
which helps to explain the relationship between exercise and cardiovascular health.
A study investigating the relationship between exercise (independent variable) and improved cardiovascular health (dependent variable)
Self-esteem as mediator to academic performance & well-being
, as high academic achievement may lead to increased self-esteem
communication as a mediator of relationship satisfaction and longevity
-> good communication fosters deeper connection btwn partners
m
"e"
diator -> how the 2 variables are r
"e"
lated -> in-between a divorce
what is in-between that might explain ->
triangulation
is a term that you might use to explain this -> it is when you get from here to here through something else
CJ ->
mediator is like a light switch
, either it is "on" or "off" and explains "why"
there is a relationship
Independent variable (IV)
-> the
basis of your groups
that you are comparing
I
manipulate
the independent variable -> (e.g. male vs female, educated vs. non-educated, etc.)
"time" is an IV
Moderator/ Interaction variable
-> influences the strength or direction of the relationship between two other variables
Social supports as a moderator
btwn stress and job performance
education level as a moderator
btwn income earned and job satisfaction
Age as a moderator
of the relationship btwn technology usage and anxiety
Moderator ->
I notice the “o”
which reminds me m
"o"
derators tell me how str
o
ng a relationship is and what direction the relationship may go*
C
o
nn
o
r McGreg
o
r is the "moderator" in MMA
pointing = direction
CJ->
moderator is like a
"dimmer switch"
-> explains why there is a relationship in the first place
it is a 3rd variable that affects the nature of the relationship
Hack for IV's & DV's
I want to know if (blank), depends on (blank)
first blank = DV
second blank = IV
Extraneous Variables (a.k.a. Confounding or Disturbance variable)
-> variables other than the IV's that cause changes to the DV's
These are "EXTRA things" outside of the IV
Anchor -
> Extra, extra read all about it -> " confounding is disturbed about the variables"
A confounding variable is a variable which is not of interest to the researcher that exerts a systematic effect on the DV.
Dependent variable (DV)
-> what your
outcome measures
I
measure
the dependent variable (a.k.a. outcome variable)
DV's don't have levels
survey says = final decision
Supressor Variable
-> it reduces or conceals the relationship btwn variables.
Scales of measurement
-> describes the nature of our data
any data we collect must be 1 of 4 types
first ask yourself -> how are we measuring our people?
are they getting a score or numerical value => if YES => that is interval or ratio data (e.g. male/female, ethnicity)
i).
Nominal -> (Nom = name)
Categorical data that is not numerical
(e.g. gender, race, political view)
ii).
Ordinal -> (ord = order)
-> data that must be in some kind of order & distance between items is not equal
(e.g. race results 1st, 2nd, 3rd; grades A,B,C,F; education level; customer satisfaction level using a Likert scale)
are they just being counted in one category or another => If YES => Data must be nominal or ordinal (e.g. are you voting republican or democrat)
iv).
Ratio -> think “Ration” is the amount of something
once you run out you have nothing left
you can't have -3 dogs
a zero in a Ratio scale does mean an absence of something
(e.g. income, height, weight, unemployment rate)
iii).
In"t"erval -> (distance between 2 things)
-> numbers only, and is the distance between 2 numbers which is equal, numbers can be negative
(e.g.
T
emperature &
T
ime -> 1 pm is between 2 and 3 pm -> 3 pm is not more time than 2 pm
Measures of Central Tendency
=> 3 possibilities
Median -
> is the middle value (number) that corresponds to the 50th percentile -> arranged in order from smallest to largest
(Me = median)
Malcolm in the median
-> this is always the middle value.
the median is a better measure when there are extreme scores,
or a substantial percentage of maximum scores
Mode
(most popular) -> is the most frequently occurring value or score ->
(mode is mOst b/c of the "o")
apple pie & ice-cream is most popular choice
Mean
-> is the average of the dataset
(meAn is the mathematical average)
Mean "Avril" Lavine
-> where everything she does, averages out in the end.
the mean is the
best measure of central tendency
Measures of Spread
Standard Deviation -
> the average amount we expect a point to differ or deviate from the mean
Range
= the difference btwn the highest and lowest score in your data
Distribution Curves
Normal Distribution
-> is a symmetrical bell-curve that has certain characteristics
68%
of scores fall between +1 and -1 SD
mean, mode and median are equal
95%
of scores are 2+ or -2 SD
99%
of scores fall between 3+ or -3 of the bell curve
TQ -> the shape of a Z-score distribution is identical to (or follows) the shape of the raw score distribution.
Positively skewed
-> mean has the highest value and the mode has the lowest
hint: you are right mean!!
So exit stage right, tail goes to the right!!
(low mode = p0sitive)
the always go in alphabetical order =>
me
an,
med
ian,
mo
de
Whale's tail => the tail tells the tale
Negatively skewed -> mean is the lowest
and mode is the highest
hint: Low mein = eating noodles with left hand, so noodles (tail) goes left
Leptokurtic Distribution (lept over)
-> sharper peak and flatter tails ->
scores are piled in the middle
Platykurtic distribution
-> scores more
evenly distributed
throughout the distribution
the mode is either on the top or bottom
& the
median is always in the middle
a graph of
percentile ranks is always flat or rectangular
Types of Data Presentation
Interval & Ratio "IR" data graphs
Histogram ->
used when there is a large distribution of scores
the bars are touching each other b/c there is a range
Line graphs
-> interval & ratio data
Bar graph
bars above each category w space between
hint: No bar = nominal & ordinal
x-axis is always horizontal -> is the categories or the scores (interval or ratio)
y-axis is vertical => is always frequency (# of people)
MATH STUFF
Square roots
before square rooting, convert the number into 100ths (e.g. 0.5 becomes .50) => square root of .50 is 7 b/c 7x7 = 49
you answer is always in expressed in 10ths (e.g.
Square root of .1 becomes .10 => 3 squared = 9, which is very close to 10. the answer is .3
Squaring decimals
answers are always expressed in 10ths (e.g. .6 squared become .36)
(e.g. .1 squared = .01)
Relationship in equations
when A = B/C
=> A & B will always have a direct relationship => (e.g. when A increases, so does B, and visa versa) =>
A and C will have an inverse (indirect)
relationship. As A goes up, C goes down
for the EPPP, this concept is used when calculating
the standard error of the mean formula
if SD of population increases => SEM increased
if sample size increases => SEM decreases
Data
-> is a measurement or observation
Key terms
p-values
less than .05
-> is the cut-off that something is statistically significant
the smaller the p-value the greater the confidence you have of rejecting the "null hypothesis"
Central Limit Theorem -
> as the sampling size increases, regardless of population, it will approach a normal distribution
Statistical Power
-> greater population homogeneity (not heterogeneity) is associated with greater statistical power
https://www.youtube.com/watch?v=zeJD6dqJ5lo
Test Construction
IV. Test Score Interpretation
T-scores and Z-scores & Stanines
T-scores
-> are used for smaller sample sizes or when the population standard deviation is unknown
T-scores have a mean of 50 & SD of 10
"T" for tens
these are also percentile scores
Z-scores
-> are a count of the SD -> applicable for large sample sizes with known population standard deviation
mean of zero & SD of 1
(a.k.a. Standard Deviation)
Stanines (aka-Stens)
-> "s" staying in the middle -> mid-score is "5"
hint: Stanines are in the middle at 5, everything starts in the middle of the curve
(a.k.a. Stens)
have a mean of 5 and SD of 2, that ranges is from 1-9.
IQ scores
IQ scores have a
mean of 100 & SD of 15
IQ scores are double b/c they are so smart
measured by t-scores
Percentile Ranks
-> 2, 16, 50, 84, 98
I. Item Analysis & Test Reliability
Latent trait model/ Item response theory-
it is assumed that item performance is related to the amount of the respondents latent trait ex. stats ability. Latent trait models are used to establish a uniform scale of measurement that can be applied to individual of varying ability and to test content of varying difficulty.
Discriminant analysis-
required when several IV are used to predict group membership (group membership is another way to say discrete categories)
Classical Test Theory (CCT)
-> a framework for developing and evaluating tests. How much can we trust the test?
measurement error
-> is due to random factors that affect the test performance of examines in unpredictable ways
examine fatigue
ambiguously worded test items
distractions during testing
true score variability
-> is the result of actual differences among examines with regard to whatever the test is measuring.
a.k.a. "True Test score theory"
Methods for Estimating Reliability
Internal Consistency Reliability
-> reliability of scores over different test items ->
best for tests that measure a single content domain
OR
aspect of a bx
b).
Kuder Richardson (KR-20)
-> it quantifies the extent to which items in a scale/ test are correlated with each other OR
the degree to which the items measure the same underlying construct
The
coefficient alpha
ranges from 0 to 1
A value closer to 0 indicates low internal consistency
-> where items do not consistently measure the same underlying construct.
A value close to 1 indicates high internal consistency,
-> where the items in the scale are highly correlated with each other.
(KR-20) is used for
dichotomous items
(e.g. yes or no, correct or incorrect)
Kuder/K.D Lang is a "dyke"
(dichotomous)
a).
Coefficient alpha (a.k.a. Cronbach's alpha)
-> it involves administering the test to a sample of examines
calculates the average
inter-item consistency
d).
Spearman-Brown prophecy formula
-> used to compensate for lengthening or shortening a test on its reliability coefficient.
should I keep my name, or should I hypenate it? -> lengthen of shorten
c).
Split-half reliability
-> is taking the test and divide it into 2 halves -> you should get the same score on both halves of the test
splitting the test in half (e.g. even- and odd-numbered items) and then correlating the 2 set of scores
A problem with split-half reliability is that the shorter tests tend to be less reliable than longer tests -> therefore a split-half reliability coefficient underestimates a test's reliability
need a mnemonic to remember the differences with these different types and how to tell them apart ...talk to Roza
2.
Alternate Forms Reliability -
> two different test are used to test the individual for the same thing & there is a passage of time between tests
Test-Retest Reliability
-> Consistency of scores over time -> you take a test, and then you re-take it, you should get similar scores if test is reliable
(
Equivalent form test -> EPPP/SAT exam
b/c you can take different tests but get similar results)
best for speed tests
4.
Inter-Rater Reliability
-> is used to determine consistency of scores when there are multiple raters
Cohen's kappa
(kappa = couple, 2 or more) -> is used to correct for chance agreement by different raters
Anchor Story Mr. Reliable Construction business ->
his motto is almost always reliable
except for 1 "measuring incident"
ii).Afro Jack
(Alternate forms)
2 different test to see which one works better -> test 1, test 2??
iii). ICR "Iceman"
(Internal-consistency-rater)
-> I just want to play beach volleyball with my "teammates"
d).
Spearman-Brown
-> compensate for long tests OR small test
a).
Cronbach (coefficient alpha)
-> this dude is consistent!!
b).
Kuder-Richardson
(KR-20) "Kuder" -> is always "k"omparing
c).
Split-half
-> always undecided
i). Testy Mctesterson
(test-retest)
I love to take tests!!
kappa chance on me
iv). Inter-rater
(Smooth operator)
-> smooth interrater....
when there is multiple raters
these are all different types of reliability
coefficients
or measures used
instead of using "reliability coefficients" use the term
RELIABILITY MEASURES
Factors that Affect the Reliability Coefficient
:
Guessing
Range of Scores
Content Homogeneity:
Reliability Index
Item Analysis
Item Difficulty
Item Discrimination
Standard Error of Measurement & Confidence Intervals
Item Response Theory
II. Test Validity - Content and Construct Validity
Multitrait-multimethod Matrix
-> is a table of correlation coefficients that provides info about a test's reliability and convergent and divergent validity
(a)
Monotrait-Monomethod Coefficient:
The monotrait-monomethod
(same trait, same method)
coefficient is a reliability coefficient (e.g., coefficient alpha) for the self-report
(b)
Monotrait-Heteromethod Coefficient
: The monotrait-heteromethod
(same trait, different method)
coefficient is the correlation coefficient for the self-report sociability test and the teacher report sociability test. When this coefficient is large, it provides evidence of the self-report sociability test's convergent validity.
(c)
Heterotrait-Monomethod Coefficient
: The heterotrait-monomethod
(different trait, same method)
coefficient is the correlation coefficient for the self-report sociability test and the self-report impulsivity test. When this coefficient is small, it provides evidence of the self-report sociability test's divergent validity.
(d)
Heterotrait-Heteromethod Coefficient:
The heterotrait-heteromethod (different trait, different method) coefficient is the correlation coefficient for the self-report sociability test and the teacher report impulsivity test. When this coefficient is small, it provides evidence of the self-report sociability test's divergent validity.
For example, assume that you want to use the multitrait-multimethod matrix to assess the convergent and divergent validity of a newly developed self-report sociability test for middle-school students, and you know that the research has found that sociability is not correlated with impulsivity. Consequently, you administer the self-report sociability test to a sample of students along with tests of sociability and impulsivity that have already been validated: a teacher report sociability test, a self-report impulsivity test, and a teacher report impulsivity test.
Convergent validity exists
, when we have high correlations of monotrait (MT) heteromethod (HM)
Divergent validity (aka discriminant validity) -> exists
when we have high correlations of heterotrait (HT) monomethod (MM)
Monotrait
= mono (1)+ trait -> personality, characteristic or bx
Heteromethod
= we are using 2 different test to measure something (e.g. Beck depression & Cam's depression test)
Heterotrait
= hetero (2) + trait -> personality, characteristic or bx
Black & white -> means CJ (coach Jeremy approved discussion)
Monomethod
= we are using 1 test to measure something
Content Validity ???
Construct Validity ???
III. Test Validity - Criterion-Related Validity
sensitivity
is "true positives" -> they have the disorder
my common sense confirms that I have it!!
specificity
is "true negatives" -> they don't have the disorder
CONCURRENT & PREDICTIVE VALIDITY
Clinical Utility
-> refers to the extent to which a test is useful
Incremental Validity
-> refers to the increase in the accuracy of predictions about criterion performance that occurs by adding a new predictor to the current methods used to make predictions.
need to improve my knowledge on test construction
III.
TYPES OF RESEARCH
-> Single Subject & Group Design
Quantitative Research
-> 3 types;
ii).
Correlational Research
-> correlating scores and comparing 2 or more variables to determine the relationship btwn the variables
Negative Correlation
Zero Correalations
Positive Correlations
i).
Descriptive Research
-> measures a set of variables a they exist naturally
Qualitative Research
-> 4 types
c).
Phenomenology
-> phenom
"lived experience"
(e.g. divorced, deployed, new job, etc.)
from the Greek word phainomenon
appearance
-> appears to be, "to show"
b).
Ethnography -
> describing a culture or community by observing them in their
"natural environment"
not interacting
(e.g. anthropologist going to a village and watching)
a).
Grounded theory
-> look at the ground ->
interviews & observations
(e.g. grounded while making observations)
collecting of data based on your hypothesis
d).
Thematic Analysis
-> describe what is going on through
images/pictures
(e.g. picture books have depth)
in-depth interview & focus groups
other type ->
Triangulation
-> comparison and combining different sources of evidence (Denzin) ; 4 types
data triangulation
-> same method at different times, different setting, different people
3.
investor triangulation
-> 2 or more investigators collect and analyze data
methodological triangulation
-> uses multiple methods (e.g. interviews, focus groups, etc.
theory triangulation
-> interpreting data using multiple theories, hypotheses or perspectives
Experiential research
-> conducted to determine if there's as
causal relationship btwn IV & DV
a). Experimental Research
Single Subject Design (A.K.A. - idiographic)
-> 5 ways
Reversal Design ABAB or ABA Design
-> where the second "A" is return to baseline and "B" is return to treatment
A common threat is that the measures may
fail to return to baseline
b/c they have already been exposed to tx.
(e.g. hyperactive child is impulsive) -> A phase is how many times he jumps out of his seat -> B is intervention
ABAB removes history as a threat
withdrawing tx when it is working is not ideal
recall strategy needed here
Multiple Baseline Design
-> when you are doing the something 3 times
Tx is applied consecutively, sequentially, or successively
mult-i-ple = 3X
An advantage of multiple-base line design over ABAB is that y
ou don't have to withdraw tx
which could be detrimental, especially if it is working
recall strategy needed here
AB Design
-> one subject, where "A" is baseline and "B" is treatment
simplest design -> Little Abner all by himself
same person is measured many times
autocorrelation
is often associated with this type of research
biggest threat is history
-> something happens at the same time that threatens our research results
Simultaneous treatment
-> looking at (2) different treatments b/c you want to know which one is better
(e.g. two types of tx at different times of the day)
probably won't see this on the exam
Idiographic
-> research that focuses on understanding the unique experiences of individuals or specific cases.
In experimental research there is a "causal" relationship btwn IV & DV
Iggy Pop is a "single idiot"
In this type of research you study 1 or few subjects in a very intense manner
single subject = is an experiment
Experimental Group Design Research (aka Nonmothetic)
c.
Mixed Design
-> is a mixture of between groups and within groups
(e.g. one component is repeated and another component is independent)
1 grouping must be independent and 1 thing is repeated measure
a.
Between Groups Design
-> comparing subjects to different groups where data and groups are independent
(hint: you MUST choose between this or the group)
data must be
independent
(can't be in more than 1 group at the same time)
Factorial Design
-> occurs whenever there is 2 or more independent variables
it allows researchers to obtain information on the the main effects of each IV.
main effect
-> is where the effect of 1 IV on the DV
interaction effect
-> the combined effect of 2 or more IV's on the DV
this is from Psychprep audios
b.
Within Subject Design
-> comparing groups when the data is repeated . Everyone has data in each one of the groups
(e.g. before, after or during treatment)
data is correlated
measured repeatedly over time
Butterfly on Coral Reef => Correlation
Sampling Methods for Group Design
-> how researcher go about selection target populations for a study; 2 types
Non-probability Sampling
-> members don't have an equal chance of being selected.
b).
Voluntary Random Sampling
-> the sample consists of individuals who volunteered to participate in the study
c).
Purposive Sampling (aka judgmental sampling)
-> when researchers use their judgment to select individuals who are appropriate for the purposes of their study
(e.g researcher wants to study beggars, and he visits 3 areas in the city where beggars live -> he then selects and interviews beggars)
researcher attempts to identify "target population" with a specific objective
a).
Convenience sampling
-> involves individuals who are easily accessible to the researcher
(e.g., university students are selected in various spots within the campus)
(e.g. during election, journalist on the stress asks random people who are they voting for)
d).
Snowball Sample
-> used when direct access to members of the target population is difficult.
(e.g. asking individuals who participate in the study to recommend others who might qualify)
psst!! pass it on
NON-RANDOM
Risks of non-random research
vulnerable to
sampling error (a.k.a. selection bias)
vulnerable to
sampling bias (a.k.a. systematic error)
Anchor Story Norando's Mac's Store Research Project
picking only people who come into the store ->
(convenience)
sign outside the store saying, "Ask to be a volunteer" ->
(voluntary random sampling)
gets mad, takes the mission statement, leaves the store and starts making decision on who he think would be a good subject ->
(purpose)
starts to snow, so he makes snowballs, everyone he hits comes to talk, so he tries to sell them on doing a survey
(snowball)
3).
Community-based Participatory Research (CBPR)
-> a collaborative approach to equally involves all partners in the research process
CBPR s
tarts with a research topic of importance with the purpose of combining knowledge with action
to achieve social change & improve health outcomes
9 core processes include;
(a) Recognize the community as a unit of identity
(b) Build on the community's strengths and resources
(c) Facilitate an equitable, collaborative, and power-sharing partnership during all phases of the research
(e) Integrate and achieve a balance between knowledge generation and intervention for the benefit of all partners.
(d) Foster co-learning and capacity building among all partners.
(f) Focus on public health problems of relevance to the community and emphasize an ecological approach that recognizes the multiple determinants of health.
(g) View system development as a cyclical and iterative process.
h). Disseminate research results to all partners and involve them in the dissemination process.
(¡) Understand that CBPR is a long-term process that requires a commitment to sustainability.
Probability Sampling
-> sampling is representative of the population.
b).
Systemic random sampling
-> used when a random list of all individuals in the population is available
It involves selecting every nth (e.g., 10th or 25th) individual from the list until the desired number of individuals has been selected
c).Stratified random sampling
-> used when the
population is heterogeneous
with regard to
one or more characteristics that is relevant
to the study (e.g., gender, age range, DSM diagnosis)
key requirement ->
it must ensure that each characteristic is adequately represented
in the sample.
dividing the population into subgroups (strata), based on the relevant characteristics
and selecting a random sample from each subgroup
a).
Simple random sampling
-> All members of the population have an equal chance of being selected
(e.g. using a computer-generated sample of individuals that was randomly chosen from a list of all individuals in the population)
it simply look like everyone will get an equal chance to be selected
d).Cluster random sampling
-> when it is impossible to randomly select individuals from a population b/c the population is large & b/c there are natural clusters within the population
It involves randomly selecting a sample of clusters and then either including in the study all individuals in each selected cluster or a random sample of individuals in each selected cluster.
(e.g., mid-sized cities, school districts, mental health clinics).
this one is a bit confusing ....NEED TO FIND AN EXAMPLE FOR THIS ONE
RANDOM
3 types of research
-> (Psychprep audio)
Quasi experimental design
-> has (1) I.V. and subjects are not randomly assigned
Non-experimental design
-> when you don't have any manipulation of the variables (e.g. comparing men and women income levels)
True experimental design
-> MUST HAVE at least (1) IV AND yoru subjects
must be randomly assigned to different groups
II.
INTERNAL/EXTERNAL VALIDITY
-> used to determine the usefulness of results in experimenal research
Internal Validity
-> how accurate are conclusion about your research and is there a cause-effect relationship
Threats to Internal validity
this stuff is in assessment, do not duplicate
some overlap, mmap in assessment
Threats to External Validity
this stuff in Assessment - do not duplicate
IV. CORRELATION & REGRESSION
Correlation coefficient
-> It is a number that tells you how two sets of data are related to each other
symbol (r) ranges from +1.0 (positive correlation); whereas -1.0 (negative correlation)
3 assumptions
a). relationship btwn variables is linear
b). there is an unrestricted range of scores for all variables
c). there is
homoscedasity
-> which means that the variability of criterion scores is similar for all predictor scores.
Regression Analysis
-> is used to predict something
correlation matrix
-> the closer the dots the stronger the relation
negative correlation
-> is an inverse/indirect relationship, as one goes up the other goes down
positive correlation
-> is a direct relationship btwn variables that goes in the same direction
Bivariate Correlation Coeffeccient
see cheat sheet on this eta, point biserial, etc.
Coefficient of Determination
Multivariate Correlation Techniques
canonical regression
multiple regression
discriminant function
Structural Equation Modeling (SEM) -> don't waste your time with this as per JEREMY
Concepts that I am unsure of where they are tied to (e.g. loose ends). I need to find a place to house this content
Criterion vs. Norm. Referenced Scores
(PSYCPREP)
Criterion-referenced scores
-> how a person did on some external criterion
percentage correct
pass/fail
raw scores
Norm-referenced scores
-> scores that compare us to others in a group (me vs. everyone else)
(e.g. percentile ranks, t-scores, z-scores)
Factor analysis -> is a test of structure or fit
-> It seeks to reduce complex correlations to an underlying set of explanatory factors.
Orthogonal
->
Un
correlated = orthogonal (lots of words) -> The lack of correlation makes the analysis easier to interpret
Oblique
-> Correlated = fewer/shorter word
Multiple Regression
regression
is another word for
prediction
-> so we are using one thing to predict something else
multiple regression equation (MRE)
-> you have multiple predictors, used to predict your outcome (e.g. factors that affect weight loss -> calories restriction and exercise)
the key thing in MRE, is the predictors are compensatory (e.g. one variable can compensate for one another)
non-compensatory is when one variable should not be used to compensate b/c they are too different from each other
